NCSU-CUSP: A Program Making a Difference in Quantitative ...

Chapter 2

NCSU-CUSP: A Program Making a Difference in Quantitative Sciences Sujit K. Ghosh

AMS Subject Classification: 62F03, 62F15, and 62P10

2.1 Introduction The Department of Statistics at North Carolina State University (NCSU) established a Computational Science Training for Undergraduates in the Mathematical Sciences (CSUMS) program funded by the National Science Foundation (NSF) under the leadership of the Principal Investigator, Professor Sujit K. Ghosh. The overarching goal is to provide a rich applied computational statistics research experience to a diverse population of undergraduate students that will encourage them to continue their academic programs to the graduate level and will help them in making more informed decisions about their academic or nonacademic careers. The NSF-CSUMS project titled NCSU Computation for Undergraduates in Statistics Program (NCSU-CUSP), prepares students to engage in a significant research experience, and to be fluent in the languages of computing, mathematics, and statistics. The program was launched on September 15, 2007, with funding from the prestigious NSF-CSUMS award and the program has been awarded a total of $770,714 to date (Award# NSF-DMS 0703392: http://www.nsf.gov/ awardsearch/showAward?AWD_ID=0703392). NCSU-CUSP targets rising senior and junior mathematics majors at NCSU and Meredith College who have demonstrated academic excellence. With rapid advances in technology, massive amounts of new data are generated daily in many scientific disciplines and the volumes are growing at a rate unprecedented in human history. For the USA to remain

S.K. Ghosh () Department of Statistics, North Carolina State University, Raleigh, NC 27695-8203, USA e-mail: [email protected] J. Rychtáˇr et al. (eds.), Topics from the 8th Annual UNCG Regional Mathematics and Statistics Conference, Springer Proceedings in Mathematics & Statistics 64, DOI 10.1007/978-1-4614-9332-7__2, © Springer Science+Business Media New York 2013

7

8

S.K. Ghosh

competitive and innovative, a diverse pool of researchers trained in novel and powerful techniques is critically needed to illustrate, model, and analyze these largesized, high-dimensional, and nonlinearly structured data. Building on resources of one of the country’s largest statistics departments, NCSU-CUSP has become one of the first computationally intensive statistics programs for undergraduates in the nation. The cutting-edge projects from this program have led to the development of new computationally intensive courses and interdisciplinary courses, which will have a long-term impact. The project is also committed at the outset to increasing diversity in the emerging field of computational statistics. NCSU-CUSP has increased awareness of statistical science among mathematics majors and faculty, it has fostered greater collaboration between interdisciplinary programs, and it has encouraged a diverse pool of wellprepared students to pursue graduate studies in quantitative sciences. To date, the program has supported 34 undergraduate students who worked in a cohort of 6–8 students in each academic year since summer of 2008. Out of these 34 students, 27 (80 %) are female out of which 2 are African-American students. Out of the 28 students who have completed the program, all of the graduating seniors have either entered into a Masters’ program or a PhD program in Statistics, and a majority of them have chosen NCSU as their graduate program. The project has supported four bright students and a faculty member from the local Meredith College, which has become one of the largest independent private women’s colleges in the United States of America (USA). A letter from Dr. E. Jacquelin Dietz (Professor and Head of the Department of Mathematics and Computer Science at Meredith College) describes the impressive impact and contribution of the NCSU-CUSP. In particular, Professor Dietz remarked “The rich experiences that (NCSU)CUSP provided them (Meredith students) in statistics, mathematics, computation and genetics will inform and inspire their teaching of future generations of young students.” The NCSU-CUSP has also supported faculty members with partial salaries (during summer) who have served as the lead instructors and mentors over the past 4 years. The instructors have expressed a strong sense of satisfaction and motivation to work with the young students. The NCSU-CUSP begins with a 10 week summer program that usually starts from the end of May through the end of July each year. During this period, three to four teams of two to three students work collaboratively with program faculty mentor. Topics explored to date range from environmental statistics (e.g., “Investigation of blood lead levels in children”) to financial statistics (e.g., “Dynamics of credit ratings”) to statistical genetics (e.g., “Optimization of Grammatical Evolution Decision Trees for detecting Epistasis” and “Comparison of analytical methods for genomic association studies”). The program couples extensive coursework throughout the academic year in computing for contemporary statistical analysis with a practicum and research lab focusing on an area of application mentioned above. Dr. Alison Motsinger-Reif has been leading the program on Statistical genetics projects for the past 2 years and in a supporting letter she succinctly summarized the broad impact of the program in making a tremendous difference with Statistics department and beyond. In particular Dr. Motsinger-Reif commented

2 NCSU-CUSP: A Program Making a Difference in Quantitative Sciences

9

“The successes of (NCSU)CUSP are clear, in many ways. The program has supported some of the most talented undergraduates at the university in performing high-quality research. . . . This (financial support) ensures gifted students from many economic background are able to participate in the program.” The NCSU-CUSP not only provided financial support to all of the enrolled students but also helped them to develop skills in data management and manipulation, converting data to a form convenient for statistical analysis, and to develop simple to complex statistical procedures and graphics. Training in communication skills helped to develop graduates who can bring scientific research results to the public and policy makers. The students benefitted from a significant, collaborative interdisciplinary scientific research experience under the mentorship of faculty working at the forefronts of their disciples. NCSU-CUSP supported all student travel allowing them to present their research work at regional, national, and international conferences. The activities of NCSU-CUSP are consistent with the recommendations of the important National Academy of Sciences’ publication “Rising Above The Gathering Storm: Engaging and Empowering America for a Brighter Economic Future.” Through this program, it has developed one of the first undergraduate-level computationally intensive and research oriented statistics curricula in the nation. In summary, the NCSU-CUSP has made a significant difference by 1. preparing undergraduate statistics/mathematics majors (in particular by engaging and encouraging women in mathematical fields) to take advantage of computing advances and make sophisticated computing an integral part of the curriculum and a significant research experience; 2. improving students’ nontechnical skills, including public speaking, written communication, ethical reasoning, and the ability to creativity in developing statistical and computing approaches to solving interdisciplinary scientific research problems and 3. preparing and motivating a diverse pool of highly qualified students to pursue interdisciplinary graduate studies in the quantitative sciences.

2.2 Program Activities and Findings There are several key aspects of the program that have lead to its successes. The program is very vertically integrated, with the program PI allowing the instructors freedom to run their cohorts in a way that best fits the subdiscipline that they are focusing on. The instructors work together as a team (there is a lecturer and a computing instructor) to teach students the skills they need for their research projects. Additionally, there have been graduate teaching assistants that have volunteered to help in the mentoring process by working with research teams. The undergraduates get the advantage of the expertise of all the team around them, and the graduate students get valuable experience in mentoring a research project. This

10

S.K. Ghosh

integration also really educates the students about the process of continuing on in academia. By working so closely with both faculty and graduate students for a full year, students get lots of opportunities to learn about life and expectation at each level. Demystifying the field helps them make a more informed decision about moving through the academic pipeline to the next step. The structure and length of the program also helps to keep the students stay in the pipeline. NCSU-CUSP is not just a summer program, its a year-long so that students have help/guidance in applying for graduate school (letters, mentoring, etc.) in their senior year. New courses have been developed in response to CSUMS activity and are open to all students. In particular, the statistical computing and data management course (which has become a required course for CSUMS students) provides a solid background on the use of computers to manage, process, and analyze data. The courses developed as a part of CSUMS activity are popular on campus and provide a strong foundation in statistics and computing needed to implement computationally intensive statistical methods. These courses have broad impacts on student training as they motivate a diverse pool of highly qualified students to pursue interdisciplinary graduate studies in the quantitative sciences. In particular, the environmental statistics practicum course has motivated students to pursue research activities in collaborations with scientists at US Environmental Protection Agency (EPA). Also the statistics and financial risk practicum course has been instrumental in motivating the students to understand and explore the mechanics of financial risk. Also, during the summer of 2010, by taking the course “Statistical Genetics Practicum,” the students learned about computer-intensive data-mining tools for gene-mapping in human genetics and explore the relative performance of these methods on both real and simulated data. The entire group of six to eight students met with their faculty advisers on a regular basis to discuss the project updates. All of them worked together on research projects, shared their research findings, collaborated on performing statistical simulations, and explored the impact of scientific theory using computational approaches. The insights gained in this collaborative exploration involving all students were then used to construct various statistical models. Computational tools included statistical modeling and data-mining software. The graduate student assigned to this project helped all eight students to learn computational methods executed via SAS. In addition, they also explored the consequences of the choice of distribution on the value of commonly used statistical metrics. All of the eight students have not only used conventional techniques as far as possible but they also used computer simulations to answer questions that these techniques cannot answer. Students were introduced to computational tools using local computing, as well as high-performance computing using NCSU’s supercomputing cluster. Additionally, all students attended field trips to see how various data are collected, as well as to super-computing facilities to see how high performing computing is made possible. Although all eight students actively collaborate with their faculty mentors on all three projects, a smaller subset of students work in teams of two on the research projects. Later a smaller subset of students took the lead in writing the manuscripts for possible publication in undergraduate research journals.


11

The Statistics department at NCSU continues the development of its web site to publicize the department’s CSUMS program: http://www.stat.ncsu.edu/cusp/. The web site is intended to blend with other departmental web sites devoted broadly to the academic, research, and human resource aspects of the department. The web site provides a detailed list of CSUMS activities including program objectives, information on financial aid, courses, research projects, and a photo gallery. The undergraduate program has a Stat Club that serves both pre-professional and social/group cohesiveness functions. One of these meetings included a presentation focusing on graduate education, its benefits in expanded career opportunities, suggestions for preparing one’s self for graduate study, and information about assistantship, fellowships and traineeship and their associated benefits. Other meetings featured speakers, sometimes past graduates of the program, who discuss their experiences as statisticians, what aspects of their training were especially valuable, and the opportunities they see for future graduates of our program. The Stat Club took a trip to the Washington DC area to visit Federal agencies that employ statisticians. The group also met with the board of directors of the American Statistical Association where they discussed the opportunity for graduate study in statistics. The undergraduates were chaperoned on this trip by a post doctoral fellow.

2.3 Program Impact The program has encouraged participants to continue to graduate school, with ALL of the graduates of the program continuing to a graduate program in a quantitative area. This program has also had a direct impact on the career goals of several students who would not have attended graduate school. The program has also had an impact beyond the participants and helped expand NCSU’s course offerings in key areas. The CUSP program recruits six of these students from NCSU each year. The impact on these students is clear. ALL of the students who participate in CUSP go on to graduate programs in quantitative fields. The extensive undergraduate research training along with training in advanced computing makes CUSP students attractive candidates for masters and PhD programs. In the first two cohorts all of the participants went on to graduate programs. In later cohorts some students have not yet graduated but are planning to go on to graduate programs. Several of the students went on to the NCSU graduate program in statistics. These students have exceptional academic records and compete well among the other students in our highly competitive program. A large number of the students went on to become part of the Masters of Science in Analytics (MSA) program at NCSU. The director of this program has expressed how the computational training and understanding of how to apply statistical theory to poorly defined problems has made our students exceptional members of the MSA program. Other students went on to graduate programs around the country. Almost every CUSP student was accepted at multiple graduate institutions. CUSP achieves this impact by bringing the participants together as a cohort that works on research projects. This cohort

12

S.K. Ghosh

mentality allows the students to see role models not only in the faculty with whom they are working but also among their peers. That allows them to envision themselves working on research in graduate school that previously seemed daunting. Overall, the CUSP program helps participants realize their desire to continue into graduate school. Obviously there may be some selection bias in this result in that many of the students who are involved in CUSP are very strong students who may have gone on to graduate school in quantitative areas regardless of their participation. However, in my role as academic advisor I have seen several students who changed their long-term plans because of the CUSP programs. For example, prior to participating in CUSP a student from the 2008 to 2009 cohort had discussed her long-term career plans with me as part of our normal advising meetings. At that time she felt that she was unsure as to how she would use statistics in a career. Although she was doing well in the theoretical courses in the undergraduate program she just did not see how these courses would apply to a “real-world” problem. This sentiment is common among our students who tend to come into statistics with a desire to solve real-world problems using quantitative methods. At that time, the student felt she would probably not go on to graduate school but instead seek employment. However, the CUSP program exposed her to how the more theoretical aspects of statistics can be translated into advanced problems in statistical methods. The program also exposed her to mentors who were working with advanced methods to solve “real-world” problems. This exposure reinvigorated her interest in academic pursuits. The student went on to the MSA program and now has started on a distinguished career as an Analytical Engineer at The SAS Institute. This impact of the CUSP program is not unique to a student. We are convinced that at least one other student in each of the cohorts would not have gone on to graduate programs if they had not participated in this program. CUSP has also had an impact outside of the students who are directly involved in the program. Traditionally, we had offered a single introductory course on statistical computing. Over the last few years many students have expressed interest in taking more courses in statistical computing. As part of the CUSP program the department now offers a second course that gives students training in more advanced statistical computing methods. This course is extremely popular with many of our undergraduate students outside the CUSP program. Through this course, the CUSP program has had an impact that is much broader than the six students enrolled in the program each year.

2.4 Conclusions and Discussions Aided by rapid advances in technology, massive amounts of new data are being generated daily across multiple scientific disciplines and are growing at an exponential rate unprecedented in human history. Researchers trained in novel and powerful techniques are critically needed to illustrate, model, and analyze these large-sized, high-dimensional, and nonlinear-structured data. NCSU-CUSP has increased the


13

awareness of statistical science among minority mathematics majors and faculty, fostered greater collaboration between departments, and encouraged a diverse pool of well-prepared students to pursue graduate studies in quantitative sciences. CUSP is a model for how programs to improved undergraduate research should work. Many programs implement Research Experiences for Undergraduates (REUs). CUSP goes beyond this typical model by incorporating a cohort structure that provides a built-in support mechanism for participants. It also supplements traditional research activities with new courses that train students in methods that they can apply immediately. This cohort structure combined with curricular transformation creates a model that can make undergraduate research work elsewhere. We firmly believe that CUSP is a program that is making a real difference among the students at NCSU and can serve as a model for real transformation at other institutions. Acknowledgments This material is based upon the work supported by the National Science Foundation under grant number DMS-0703392. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author and do not necessarily reflect the views of the National Science Foundation.